Discussion Questions: 10 April

1. (JT) What is really doing the work in Logan's computational theory of apprehension - the processing or the representation? What would Brian Smith say about Logan's use of representation in this model?
2. (JT) How would Logan capture Mike's example of "pears in a bowl" with a spatial template? Would you expect his computational model to apprehend the same spatial relation for the top pear in a bowl of pears and one suspended above the bowl? Is Logan's "spatial template" theory missing anything and if so, could it be augmented?
3. (JT) Logan claims basic relations are more "complex computationally" than deictic relations, and deictic relations are more "complex computationally" than intrinsic relations. Does this contradict with his results reported in 13.5.2 (page 504)? What does Logan mean by "complex computationally?
4. (CP) It's interesting to see our usage of words like 'on' and 'over' in Experiment 1. Some languages, which might have more words for different types of On, would probably have different, more uniform results. Do you think English (or any language) is impoverished because of this, or is it transparent?
5. (JN) Are there specific experiments which might confirm or falsify the hypothesis of spatial frames underlying locative words?
6. (JN) The process of using spatial frames seems to center on "choice" -- selecting the frame that fits best from among all possibilities and then applying its associated word. Has this returned us to the school of thought in which all AI is searching? If so, where have we been in the meantime?
7. (MG) What sorts of predictions about acquisition of spatial language could we make from Logan and Sadler's model?